Welcome to the MacNN Forums.

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

You are here: MacNN Forums > Software - Troubleshooting and Discussion > macOS > Quartz accel comments from an Apple emp

Quartz accel comments from an Apple emp
Thread Tools
JB72
Mac Elite
Join Date: Jan 2001
Location: L.A., CA
Status: Offline
Reply With Quote
May 9, 2001, 05:43 PM
 
Taken from xlr8yourmac

Expert's Comments on 2D Quartz Acceleration in OS X: - Ken Dyke, former programmer with 3dfx and now with Apple posted some comments in the forums here about graphics card (GPU) acceleration for OS X's Quartz API. Here's a snippet:

" Just a little FYI...

The Quartz/CoreGraphics imaging model is essentially the same as PDF/PostScript, and makes heavy use of Bezier curves. There are no GPUs out there that can rasterize PostScript/PDF paths, so the CPU is at the very least stuck with the burden of rasterizing an alpha mask that represents the anti-aliased shape of the path you are trying to draw. Given such an alpha mask you might be able to use it as a texture to then render a character glyph or whatever else you're trying to draw.

QuickDraw and GDI use a very simplistic imaging model (basically just simple lines, rectangles, polygons, etc.) with no AA coverage information required. All of these are fairly straightforward to accelerate in hardware, and this has been a "solved" hardware problem for many many years. Unfortunately, this means that the "state of the art" in 2D imaging has pretty much not changed since the 80's (unless you count Display Postscript).

Also, it's probably worth mentioning that the original Macintosh didn't ship with a GPU. It took *years* for any kind of 2D accelerator to show up that was capable of doing any part of QuickDraw, and in the meantime QuickDraw had to use the CPU for *everything*. I think people have just gotten "used" to how fast the older 2D rendering APIs can run, without realizing that people have been optimizing them (and the hardware accelerators for them) for nearly ten years.

With Quartz, a huge leap in 2D image quality was raised (for better or worse), and we've once again surpased what the current generation of graphics hardware is capable of dealing with.

Over time I suspect we'll see GPUs that do more of what Quartz needs, but right now they simply don't exist.

IMHO, I'd much rather have the current PDF/PostScript based 2D rendering API than yet another tired circa-1980's style rendering API, and the speed hit doesn't bother me in the least, even on my stock G3-400 (okay, so it has 768MB of RAM...).
-Ken "

In most ways, not very good news. I wonder there isn't *something* more that will be doable with the GeForce 3?


------------------
JB71.34.1b

Dock Perverts still rule OK?

[This message has been edited by JB72 (edited 05-09-2001).]
     
Severed Hand of Skywalker
Addicted to MacNN
Join Date: Apr 2001
Location: The bottom of Cloud City
Status: Offline
Reply With Quote
May 9, 2001, 05:49 PM
 
Aww crap. That is the last thing I needed to hear.

------------------

"Ahhhhhhhhhhhhhhhh"
     
tie
Professional Poster
Join Date: Feb 2001
Status: Offline
Reply With Quote
May 10, 2001, 02:27 AM
 
Well, we were all thinking "maybe this summer" but the right time frame is actually ~10 years. Basically, we can't hope for any acceleration until Microsoft puts the same thing into their operating system.
The 4 o'clock train will be a bus.
It will depart at 20 minutes to 5.
     
JB72  (op)
Mac Elite
Join Date: Jan 2001
Location: L.A., CA
Status: Offline
Reply With Quote
May 10, 2001, 02:40 AM
 
Originally posted by tie:
Basically, we can't hope for any acceleration until Microsoft puts the same thing into their operating system.
That's what I'm afraid of. Though my delusional optimism is hoping that the "programmable" aspects of the GeForce 3 will allow Quartz acceleration in some way.

Maybe Apple will be able to add a "sub-GPU" to their version of the NVidia cards, or maybe to the motherboard itself. Then there's always Altivec. Who knows.

They're doing pretty good then, considering. Someone has to advance the state of the PC.


------------------
JB71.34.1b

Dock Perverts still rule OK?
     
Scrod
Mac Elite
Join Date: Jan 2001
Location: Sad King Billy's Monument on Hyperion
Status: Offline
Reply With Quote
May 10, 2001, 02:41 AM
 
Not unless that hardware acceleration comes from Apple. And don't think they won't, either.
I abused my signature until she cried.
     
JLannoo
Junior Member
Join Date: Apr 2001
Location: Harrison Twp. MI
Status: Offline
Reply With Quote
May 10, 2001, 03:20 AM
 
Well the things that really bother me in OS X are things like application launch times and a choice of several slow @ss web browsers. Both of which I'm confident will be improved immensely
as time goes on.

Even if we don't get the acceleration from the video cards like we hoped. There still plenty of things too that can be done to make the OS feel faster,and that's what counts. So don't give up hope.

------------------
-JLannoo
TiVo Zealot

G4 Cube 450
448MB RAGE 128 Pro

Athlon ThunderBird 800
256MB GeForce 2 GTS

[This message has been edited by JLannoo (edited 05-10-2001).]
-JLannoo
TiVo Zealot
G4 Cube 450
448MB RAGE 128 Pro
Athlon ThunderBird 800
256MB GeForce 2 GTS
     
aaroncsmith
Junior Member
Join Date: Mar 2001
Status: Offline
Reply With Quote
May 10, 2001, 03:37 AM
 
     
smacintush
Addicted to MacNN
Join Date: Sep 2000
Location: Across from the wallpaper store.
Status: Offline
Reply With Quote
May 10, 2001, 03:50 AM
 
Did you guys ever think that this may be one reason that Apple has started with the MP's? The last "rumor" I heard was that Apple wanted a quad 1 Ghz G4 by the end of this year, I think this could handle the Quartz rendering just fine.

Besides, the ONLY rendering problem that actually gets in the way is window resizing. I think this will be fixed fairly soon.

Just my 2�


------------------
Believe me, I'm always
right. Unless I'm
mistaken.
Being in debt and celebrating a lower deficit is like being on a diet and celebrating the fact you gained two pounds this week instead of five.
     
tie
Professional Poster
Join Date: Feb 2001
Status: Offline
Reply With Quote
May 10, 2001, 03:53 AM
 
Right. I'll believe that when I see it. Don't get too excited yet.
The 4 o'clock train will be a bus.
It will depart at 20 minutes to 5.
     
Cipher13
Registered User
Join Date: Apr 2000
Status: Offline
Reply With Quote
May 10, 2001, 04:24 AM
 
Well well well... whaddya know 'ey?

Cipher13

------------------
AIM: Cipher1387
ICQ: 48111606
mail: [email protected]

[This message has been edited by Cipher13 (edited 05-10-2001).]
     
clebin
Grizzled Veteran
Join Date: Oct 2000
Location: Cardiff, Wales
Status: Offline
Reply With Quote
May 10, 2001, 05:59 AM
 

Hee hee.

Sorry to all those waiting for Apple to support their 'super powerful' Rage Pro chips...

Chris
     
clebin
Grizzled Veteran
Join Date: Oct 2000
Location: Cardiff, Wales
Status: Offline
Reply With Quote
May 10, 2001, 06:01 AM
 

Hee hee.

Sorry to all those waiting for Apple to support their 'super powerful' Rage Pro chips...

Chris
     
Group51
Senior User
Join Date: Feb 2001
Location: United Kingdom
Status: Offline
Reply With Quote
May 10, 2001, 08:31 AM
 
I don't understand this. Is he saying that QuickTime can't be accelerated, or that OpenGL can't be accelerated? These are also things that are dead slow in MacOS X with Rage Pro. Not just window resizing.

     
Norm1985
Mac Elite
Join Date: Jun 2000
Location: Northbrook, IL, USA
Status: Offline
Reply With Quote
May 10, 2001, 08:34 AM
 
Actually, OpenGL was acclerated. But that's because he has much wider support by many OSes and people.

------------------
.OpenStep folder, don't leave Mac OS X booting without it.


[email protected]
AIM: Norm1985
ICQ: 34049393
     
Group51
Senior User
Join Date: Feb 2001
Location: United Kingdom
Status: Offline
Reply With Quote
May 10, 2001, 08:43 AM
 
Originally posted by Norm1985:
Actually, OpenGL was acclerated. But that's because he has much wider support by many OSes and people.

On the RagePro? Why is the screensaver so bad then, and why do OpenGL games either fail to run or just turn the screen white (forcing a reboot). (I have 256MB RAM)

     
JLL
Professional Poster
Join Date: Apr 1999
Location: Copenhagen, Denmark
Status: Offline
Reply With Quote
May 10, 2001, 09:05 AM
 
Originally posted by Group51:
On the RagePro? Why is the screensaver so bad then, and why do OpenGL games either fail to run or just turn the screen white (forcing a reboot). (I have 256MB RAM)
OpenGL is not accelerated on RagePro because there are no RagePro drivers for Mac OS X - yet (hopefully).

------------------
JLL
JLL

- My opinions may have changed, but not the fact that I am right.
     
urp
Guest
Status:
Reply With Quote
May 10, 2001, 10:24 AM
 
What are you all whinging about. MS has had alpha transparency since Win2K [turn it on, bog the system], and has moved to vectors in XP with GDI+. What does this mean? The other 95% of the GPU market is also looking for the same features.

BTW beziers are very lightweight and easy to deal with. Its the transparency and alpha masking that bogs things down.


flame me -- i'm stuped
     
Norm1985
Mac Elite
Join Date: Jun 2000
Location: Northbrook, IL, USA
Status: Offline
Reply With Quote
May 10, 2001, 10:37 AM
 
Well, at least Win2K (if that even is true) allows you to turn that sh*t off. Also, from what I've heard, it runs a lot better even with that junk on.

------------------
.OpenStep folder, don't leave Mac OS X booting without it.


[email protected]
AIM: Norm1985
ICQ: 34049393
     
foobars
Mac Elite
Join Date: Jan 2001
Location: Somewhere in the land surrouding Fenway Park
Status: Offline
Reply With Quote
May 10, 2001, 10:37 AM
 
Originally posted by urp:
Its the transparency and alpha masking that bogs things down.
Infact this is hardly the case. Certainly there is a small speed hit because of the math involved in trasnparency but the truth is Quartz really does a great job in this area. Windows drag very quickly despite drop shadows. If you're a programmer you can even turn those shadows off- I suggest doing it and comparing the speed. You'll notice windows are just as fast minus the extra trasnparency.
     
Milio
Grizzled Veteran
Join Date: Oct 2000
Status: Offline
Reply With Quote
May 10, 2001, 10:38 AM
 
Jeez, I've been saying this since Quartz was announced, but no one believed me.

Quartz with its PDF foundation is much more computationally intensive than QuickDraw and what we have today simply can't run it well.
     
Gametes
Mac Elite
Join Date: Sep 2000
Location: Norfolk, Va
Status: Offline
Reply With Quote
May 10, 2001, 10:38 AM
 
Exactly.
If you look to the specific features the Apple employee above stated as needing to be moved to the GPU -- which we've all known for some time -- it's not too difficult to see that these kinds of graphic actions aren't obscure enough to be "Mac only".
I'm sure many games would love to make better use of Bezier curves and transparency, meaning that these sorts of actions will make their way probably fairly rapidly into the next iteration (not even generation) of video chips.
I'm not all that worried about Quartz.

------------------
you are not your signature
you are not your signature
     
Mr K
Junior Member
Join Date: Oct 2000
Location: Australia
Status: Offline
Reply With Quote
May 10, 2001, 10:40 AM
 
Anybody technical out there know the answer?

We realize that there are limiting factors in how many G4's you can cram in a box before it makes no significant difference to the processing speed. I'm not exactly sure why that is...to do with access to disks etc?

well, of most of the graphics has to be handled by the system, would this then easily be handled by the extra G4?

So, say 2 G4's is sort of twice as fast for some proccesses. 4 is faster, but in decreasing gains in speed. But one of the processes, or a bit of each one is handling the graphics, so this may not need access to the hard disk etc etc - which means if you've got more chips than the graphics are fine.

any ideas?

K
     
Milio
Grizzled Veteran
Join Date: Oct 2000
Status: Offline
Reply With Quote
May 10, 2001, 11:02 AM
 
Originally posted by Mr K:
well, of most of the graphics has to be handled by the system, would this then easily be handled by the extra G4?
The problem is that the G4 (or P4 or Athlon) is not a graphics chip, it's a CPU. It's designed to be a general-purpose processor. Think about this, the Radeon has 30 million transistors compared to the G4 with 10.5 million, and the Radeon's sole purpose is to make graphics go faster. Now which would you rather have working on accelerating Quartz, another G4 or something like the Radeon?

Edit: I'm just gonna keep adding to this...

FYI, the GeForce3 has 57 million transistors.

[This message has been edited by Milio (edited 05-10-2001).]
     
muchfresh
Mac Enthusiast
Join Date: Jan 2001
Location: ny ny usa
Status: Offline
Reply With Quote
May 10, 2001, 11:03 AM
 
Quartz ma be not be hardware accelerated but it aint the problem

repeat after me:
"Quartz is not the problem"

The speed issues that everyone complains about are not adversly affected by quartz:
1) menu hesitation
2) window resizing
3) application launch time
4) finder navigation

Quartz is surprisingly fast.
1) scrolling is fast
2) resising stickies windows is fast
3) transparent terminal window drags fast
4) window dragging is fast
5) PDF files scroll much faster then top line W2K

The big problem is CarbonEvents. Quartz can draw pixels plenty fast. OSX just can't communicate events from process to process very fast.
'Satisfy the urge and discover the need' Q-Tip
     
muchfresh
Mac Enthusiast
Join Date: Jan 2001
Location: ny ny usa
Status: Offline
Reply With Quote
May 10, 2001, 11:07 AM
 
Flash is a perfect example of what I am suggesting.

1) view a large fash animation in IE on OSX. no problem
2) play a flash game in IE on OSX. Slooooow

its not a pixel drawing problem its the event system.
'Satisfy the urge and discover the need' Q-Tip
     
foobars
Mac Elite
Join Date: Jan 2001
Location: Somewhere in the land surrouding Fenway Park
Status: Offline
Reply With Quote
May 10, 2001, 11:11 AM
 
Originally posted by muchfresh:
The big problem is CarbonEvents. Quartz can draw pixels plenty fast. OSX just can't communicate events from process to process very fast.
Good post- I agree with you. I would also point out that process prioritizing is not working right now at all- so in a roundabout way messaging in general is delayed. Hopefully soon Darwin will allow up to nice commands and the Window Manager will gain priority!
     
Milio
Grizzled Veteran
Join Date: Oct 2000
Status: Offline
Reply With Quote
May 10, 2001, 11:12 AM
 
Originally posted by muchfresh:
Quartz ma be not be hardware accelerated but it aint the problem

repeat after me:
"Quartz is not the problem"
Quartz is a problem. Or more accurately, the inability of current hardware to accelerate Quartz is a problem. There are other things you also mention that are not Quartz related that are also problematic.

As for your examples of why Quartz is supposedly not slow, different apps will handle drawing and resizing and such slightly differently. Quartz is an amazing graphics architecture. And in many cases it doesn't get in the way too significantly. It also helps that we currently aren't asking much of our systems since there aren't many apps yet. So Stickies can take more CPU time right now because we aren't really taxing the CPU much. But imagine how much faster the system will be when Quartz doesn't hit the CPU at all (relatively speaking.)
     
foobars
Mac Elite
Join Date: Jan 2001
Location: Somewhere in the land surrouding Fenway Park
Status: Offline
Reply With Quote
May 10, 2001, 11:21 AM
 
Originally posted by Milio:
Quartz is a problem. Or more accurately, the inability of current hardware to accelerate Quartz is a problem. There are other things you also mention that are not Quartz related that are also problematic.

As for your examples of why Quartz is supposedly not slow, different apps will handle drawing and resizing and such slightly differently. Quartz is an amazing graphics architecture. And in many cases it doesn't get in the way too significantly. It also helps that we currently aren't asking much of our systems since there aren't many apps yet. So Stickies can take more CPU time right now because we aren't really taxing the CPU much. But imagine how much faster the system will be when Quartz doesn't hit the CPU at all (relatively speaking.)
I don't think you will be happy to see the performance of a GPU with only 16 (maybe 32) MB of RAM when compared to a G4 with then entire system memory. Right now getting Quartz to run fast on next-generation cards is not as an important an issue as speeding up the Window Manager and system messaging right now for current models.

And if you want to start a conversation about seperate chips handling Quartz, I suggest you take a look at Apple purchace of Raycer and decuct what you will from that...

And I just because there aren't many DOESN'T mean people aren't taxing thier system- that is a silly statement...

[This message has been edited by foobars (edited 05-10-2001).]
     
Milio
Grizzled Veteran
Join Date: Oct 2000
Status: Offline
Reply With Quote
May 10, 2001, 11:32 AM
 
Originally posted by foobars:
I don't think you will be happy to see the performance of a GPU with only 16 (maybe 32) MB of RAM when compared to a G4 with then entire system memory. Right now getting Quartz to run fast on next-generation cards is not as an important an issue as speeding up the Window Manager and system messaging right now for current models.

And if you want to start a conversation about seperate chips handling Quartz, I suggest you take a look at Apple purchace of Raycer and decuct what you will from that...
I don't disagree that the Window Manager and system messaging may probably be a higher priority. I'm simply pointing out that Quartz has it's own share of problems as well.

As for GPU vs CPU, I am a firm believer from experience in offloading intensive tasks from the CPU to a dedicated processor. A graphics card like the Radeon may only have 16MB of ram, but the entire architecture is dedicated to accelerating graphics. As for Raycer, I don't care to speculate on vaporware. Quartz is here, Radeon and GeForce3 is here, let's look at how we can get the most out of what we have today.
     
snodman
Mac Enthusiast
Join Date: Oct 1999
Location: Poway, CA USA
Status: Offline
Reply With Quote
May 10, 2001, 12:34 PM
 
Remember about a year ago Apple bought a small graphics chip maker. One of the BIG ideas about OS X (from Apple's standpoint) is to sell more NEW Macs. My guess is that the next generation motherboards HAVE a Quartz accelerator in silicon on the motherboard. This would let even a cheapie "iMac II" (or whatever they call the new box to replace the iMac this summer) run OS X faster even with G3 CPUs.
     
anand1234
Guest
Status:
Reply With Quote
May 10, 2001, 01:01 PM
 
Thank you snodman. Apple will put a graphics chip on board to speed things up. And it will be there on the new macs in July. Dorsal has mentioned this on the appleinsider boards and it makes sence. OS X appears to be too slow for some people. To spead things up, they are going to have to do something.
     
GnOm
Mac Enthusiast
Join Date: Apr 2001
Location: Earth?
Status: Offline
Reply With Quote
May 10, 2001, 01:39 PM
 
Originally posted by snodman:
My guess is that the next generation motherboards HAVE a Quartz accelerator in silicon on the motherboard.

while this _may_ be true (I�ll believe it when I hold one of those machines in my own hands), what about the rest of us? I don�t have another 200bucks to spend for a Quartz HW accellerator card, even if I had there�s no PCI Slot in my Cube...


cu
     
muchfresh
Mac Enthusiast
Join Date: Jan 2001
Location: ny ny usa
Status: Offline
Reply With Quote
May 10, 2001, 02:07 PM
 
Jimminy Crickets Folks!!

Quartz may be not be hardware accelerated but it aint the problem
repeat after me:
"Quartz is not the problem"

The speed issues that everyone complains about are not adversly affected by quartz:
1) menu hesitation
2) window resizing
3) application launch time
4) finder navigation

Quartz is surprisingly fast.
1) scrolling is fast
2) resising stickies windows is fast
3) transparent terminal window drags fast
4) window dragging is fast
5) PDF files scroll much faster then top line W2K

The big problem is CarbonEvents. Quartz can draw pixels plenty fast. OSX just can't communicate events from process to process very fast
Flash is a perfect example of what I am suggesting.
1) view a large fash animation in IE on OSX. no problem
2) play a flash game in IE on OSX. Slooooow

its not a pixel drawing problem its the event system.
'Satisfy the urge and discover the need' Q-Tip
     
SkullMacPN
Mac Enthusiast
Join Date: Feb 2001
Location: Savannah, GA
Status: Offline
Reply With Quote
May 10, 2001, 03:17 PM
 
Originally posted by muchfresh:
Quartz ma be not be hardware accelerated but it aint the problem

repeat after me:
"Quartz is not the problem"

The speed issues that everyone complains about are not adversly affected by quartz:
1) menu hesitation
2) window resizing
3) application launch time
4) finder navigation

Quartz is surprisingly fast.
1) scrolling is fast
2) resising stickies windows is fast
3) transparent terminal window drags fast
4) window dragging is fast
5) PDF files scroll much faster then top line W2K

The big problem is CarbonEvents. Quartz can draw pixels plenty fast. OSX just can't communicate events from process to process very fast.
Not Quartz's Problem?
1) No, that is a quartz problem. The transparency of the menu has to be calculated by the CPU before the menu is displayed. And, worse yet, the fade out has to be calculated as/before the next menu is calculated if you scrub across the menus.
2) Somewhat... the dropshadow is calculated, but the rest is (mostly) a developer problem.
3) Correct.
4) Correct.

Quartz is fast? Yes but...
1) That's hardware accelerated... just like any other OS.
2) Your right, window resizing is a matter of developer savvy and implementation. If a developer blocks content update in windows with lots of data (like TextEdit will do), then it goes reasonably fast.
3) See #4, but a transparent terminal does drag slower than a sold one.
4) Yes, because the action of simply moving pixels around the screen (or what equates to moving data around in VRAM) is fast, especially when everything behind the window is buffered. The only thing calculated here is the shadow around the window (and the title bar, because you can drag a background window).
5) Yes, because the whole graphics subsystem is based on PDF, the routines for rendering vector graphics are very fast.

It's been said before, and I'll say it again. Quartz is VERY fast for what it does (which is why it beats the pants off of any other OS in PDF opperations). However, all the code optimization in the world isn't going bring Quartz up to the speed of a fully hardware accelerated display system (like QuickDraw) untill there is more robust hardware acceleration. Hardware acceleration will come. Will it come in the form of new hardware? Perhaps using the 3D GPU to calculate transparency? I don't know the answer to that one... but it'll come.
     
sCreeD
Fresh-Faced Recruit
Join Date: Sep 2000
Status: Offline
Reply With Quote
May 10, 2001, 03:22 PM
 
Ah yes, the mythical Raycer chipset.

Originally posted by GnOm:

while this _may_ be true (I�ll believe it when I hold one of those machines in my own hands), what about the rest of us? I don�t have another 200bucks to spend for a Quartz HW accellerator card, even if I had there�s no PCI Slot in my Cube...


cu
Um, no. No-no-no, Apple doesn't do PCI upgrade cards; they sell Macs -- the more Macs the better. If it exists, it will be an integrated chipset on new Macs (UMA 2.0 perhaps??).

What about "older" Macs?

I'll repeat myself, Apple sells Macs -- the more Macs the better. Apple's loyalty to their established base doesn't breach that rule.

Screed
     
JB72  (op)
Mac Elite
Join Date: Jan 2001
Location: L.A., CA
Status: Offline
Reply With Quote
May 10, 2001, 04:23 PM
 
Originally posted by anand1234:
Apple will put a graphics chip on board to speed things up. And it will be there on the new macs in July. Dorsal has mentioned this on the appleinsider boards
I must have missed that post.



------------------
JB71.34.1b

Dock Perverts still rule OK?
     
Skywalkers new Hand
Grizzled Veteran
Join Date: May 2001
Location: At the end of Lukes Arm.
Status: Offline
Reply With Quote
May 10, 2001, 04:45 PM
 
"Apple will put a graphics chip on board to speed things up. And it will be there on the new macs in July. Dorsal has mentioned this on the appleinsider boards"

The why the HELL is it not on the BRAND ****ING NEW iBooks! That really gets me mad.

------------------

"Wedge, pull out! You're not doing any good back there!"
     
gorgonzola
Admin Emeritus
Join Date: Nov 2000
Location: New Yawk
Status: Offline
Reply With Quote
May 10, 2001, 04:45 PM
 
Keep this in perspective. While this is a major downer, all this means is that OS X will never be as fast as OS 9 on current hardware; there are significant improvements to be made to the CoreOS that will at least improve the current situation, if not quite to the extent we had hoped.

A build of OS X with a fixed Mach scheduler should help a great deal, for example.

------------------
the oddball newsletter
------------------
it's only after you lose everything that you're free to do anything
"Do not be too positive about things. You may be in error." (C. F. Lawlor, The Mixicologist)
     
GnOm
Mac Enthusiast
Join Date: Apr 2001
Location: Earth?
Status: Offline
Reply With Quote
May 11, 2001, 01:41 AM
 
Originally posted by sCreeD:
Um, no. No-no-no, Apple doesn't do PCI upgrade cards; they sell Macs

darn, I almost forgot


cu
     
isidor
Fresh-Faced Recruit
Join Date: Apr 2001
Status: Offline
Reply With Quote
May 11, 2001, 06:08 AM
 
Basically Dyke says "we knew Quartz would be slow but we didn't want to use another kind of graphics layer because fast graphics are soooooo 90's. People think oSX is slow just because they are used to fast graphics"

Now I'm very worried.
     
JB72  (op)
Mac Elite
Join Date: Jan 2001
Location: L.A., CA
Status: Offline
Reply With Quote
May 11, 2001, 06:19 AM
 
Originally posted by gorgonzola:
Keep this in perspective... there are significant improvements to be made to the CoreOS that will at least improve the current situation, if not quite to the extent we had hoped.

A build of OS X with a fixed Mach scheduler should help a great deal, for example.
Good call. There IS other things that will speed up the OS (hardware will get faster, Darwin and OS X will mature.) Apple is the company to make the jump to the next level. I just hope they will support it with encouraging hardware to be developed to accelerate Quartz. I just like things super snappy, and great looking. Typical Mac user I am. I want it all, with a minimum premium .




------------------
JB71.34.1b

Dock Perverts still rule OK?
     
jamesa
Grizzled Veteran
Join Date: Sep 2000
Location: .au
Status: Offline
Reply With Quote
May 11, 2001, 08:15 AM
 
Originally posted by Mr K:
Anybody technical out there know the answer?

We realize that there are limiting factors in how many G4's you can cram in a box before it makes no significant difference to the processing speed. I'm not exactly sure why that is...to do with access to disks etc?

well, of most of the graphics has to be handled by the system, would this then easily be handled by the extra G4?

So, say 2 G4's is sort of twice as fast for some proccesses. 4 is faster, but in decreasing gains in speed. But one of the processes, or a bit of each one is handling the graphics, so this may not need access to the hard disk etc etc - which means if you've got more chips than the graphics are fine.

any ideas?

K
I'm not exactly sure what you're asking, but if Apple standardises on multi-CPU systems (what I'm hoping) it will provide significantly greater encouragement for vendors to thread their applications decently - and then the speed boost for multi-cpu machines will be enormous. No longer will it be a case of who wins the mhz war - it's who can get the most bang for the buck chips into their machines. Let me give you this example (prices may be slightly out of date)

I'll use Intel for this, because I don't have as easy access to pricing for MOT and IBM chips.
http://shop2.outpost.com/product/74734/ (I'm sure everyone could find cheaper prices, but this'll do for the time being)
a) P4 1500mhz - $899 US
b) P4 1300mhz - $489 US
c) P3 933mhz - $269 US

Now, of course Intel will be making a larger margin on their bleeding edge chips. But so too will MOT and IBM.

Now, for 2 P4 1500's, I can get 3.67 P4 1300's (yeah, I know, you can't get 2/3 of a chip but it's not much more to get the last 1/3).
And for those 2 P4 1500's, I can get 6.69 P3 933's (as above for previous parentheses).

Assume for a minute that Pentiums can be directly comparable by clock speed - which much more true than Pentiums vs PPCs - and you get for your $1798 either

a) 3000mhz
b) 4779.95mhz
c) 6236.18mhz

As you can see, the advantages of cheap MPs become clear. Obviously, this fails to take into account the fact that scaling isn't a linear performance improvement, etc etc. But I think you could make a pretty good argument - extracting from those Spec tests for MP, there seemed to be a 25% hit for 2 processors of 75mhz vs 1 processor of 150mhz. But weigh in the economics to those spec tests (assuming they hold true above) you'll still get significantly more bang for the buck on one of the latter options.
note: I cannabalised part of this from an earlier post to a mailing list, discussing the virtues of multiprocessing machines. The spec tests refer to a SPARC system; a 150mhz SPARC quite convincingly beat a dual 75mhz one. Other than the fact that we're not comparing 2 500mhz G4s to a 1ghz one, the G4 has had significant optimisations made to it so that it runs well in a multi-CPU configuration.

I think the two big obstacles are getting programmers to think more in a multi-threaded mindset, and getting consumers to believe in The Joy that is Symmetric Multiprocessing.

And it's for very similar reasons Apple need to ship iMacs with G4s - the number of vendors taking advantage of AltiVec is nowhere near what it should be, because the Altivec market is so small.

-- james


------------------
"Note: Silencing the alarm does not solve the problem that caused it."
-- Sola (UPS) Users Guide

[This message has been edited by jamesa (edited 05-11-2001).]
     
Group51
Senior User
Join Date: Feb 2001
Location: United Kingdom
Status: Offline
Reply With Quote
May 11, 2001, 08:22 AM
 
Originally posted by gorgonzola:
Keep this in perspective. While this is a major downer, all this means is that OS X will never be as fast as OS 9 on current hardware; there are significant improvements to be made to the CoreOS that will at least improve the current situation, if not quite to the extent we had hoped.

A build of OS X with a fixed Mach scheduler should help a great deal, for example.

Quartz, QuickTime and OpenGL. Are these all the same things? If the current cards can't do much with Quartz, can they accelerate the other two?
     
jamesa
Grizzled Veteran
Join Date: Sep 2000
Location: .au
Status: Offline
Reply With Quote
May 11, 2001, 08:40 AM
 
Originally posted by Skywalkers new Hand:
"Apple will put a graphics chip on board to speed things up. And it will be there on the new macs in July. Dorsal has mentioned this on the appleinsider boards"

The why the HELL is it not on the BRAND ****ING NEW iBooks! That really gets me mad.
I'll tell you why it's not on the brand new iBooks - because it's not going to happen. There is no way that Apple is going to add another GPU - they already have ATi's on everything when the GPU is on the motherboard. The very notion that they're going to come out with some custom graphics card is ridiculous. Do you think Apple could produce a proprietary GPU (read: small market, no economies of scale) and have it compete with the price-performance of ATi and nVidia?

There's no way it's going to happen. Apple is not SGI.
     
Spheric Harlot
Clinically Insane
Join Date: Nov 1999
Location: 888500128, C3, 2nd soft.
Status: Offline
Reply With Quote
May 11, 2001, 09:00 AM
 
Originally posted by isidor:
Basically Dyke says "we knew Quartz would be slow but we didn't want to use another kind of graphics layer because fast graphics are soooooo 90's. People think oSX is slow just because they are used to fast graphics"

Now I'm very worried.
Urm...

Basically Dyke says "we knew Quartz would be slower than OS 9 on identical hardware, but we didn't want to use another kind of graphics layer because pixel-based graphics are a complete dead end (especially with higher-resolution displays), and Quartz just gives us soooo many possibilities for the future that we wouldn't have otherwise. People think oSX is slow just because it isn't done yet, and because they don't remember what OS 7.0 was like on the high-end hardware of the time."

Of course, we could just go back to text-based interfacing - now THAT's fast! ...oh wait, we have...and the Terminal seems to zip along just fine, too...



-chris.

------------------
...the original Genie Junkie�

[This message has been edited by Spheric Harlot (edited 05-11-2001).]
     
muchfresh
Mac Enthusiast
Join Date: Jan 2001
Location: ny ny usa
Status: Offline
Reply With Quote
May 11, 2001, 09:32 AM
 
Folks this is getting tired:

**Quartz is not the problem**

say it again

**Quartz is not the problem**

OK now look at the facts
1) window dragging, fast
2) scrolling, fast
3) Stickies window resizing, fast
4) dragging transparent terminal window, fast
5) Flash animations, no problems

Quartz is very fast at painting pixels to the screan. The problem is much more fundemental. One that hardware is not going to fix anytime soon. I believe that CarbonEvents is the problem. OSX has major perfformance issues when passing messages between processes.

If you think Quartz is slow please tell me why.
'Satisfy the urge and discover the need' Q-Tip
     
Milio
Grizzled Veteran
Join Date: Oct 2000
Status: Offline
Reply With Quote
May 11, 2001, 09:51 AM
 

Originally posted by muchfresh:

If you think Quartz is slow please tell me why.
I have posted the technical reasons to this forum several times over the past couple years. Unfortunately the posts seem to have disappeared in a black hole of time.

Suffice it to say that a vector based system like Quartz takes more complex math and more computing cycles than the basic primitives and raster graphics of QuickDraw. That means slower.

Now Quartz is amazing for what it does and how quickly it does it on the given hardware. But because of what it is capable of it is slower than QuickDraw.

As has also been pointed out, QuickDraw is accelerated in the graphic cards, taking that drain off the CPU. Quartz does not yet have this luxury. So while the CPU is being asked to do the CarbonEvents you want improved, it's also trying to do all the calculations for Quartz. Yet another reason why Quartz is effectively slower than QuickDraw.

Originally posted by muchfresh:
**Quartz is not the problem**

say it again

**Quartz is not the problem**
There is not just one problem. Quartz is currently a problem. CarbonEvents appears to be another. The effects are cumulative, not exclusive.

[This message has been edited by Milio (edited 05-11-2001).]
     
muchfresh
Mac Enthusiast
Join Date: Jan 2001
Location: ny ny usa
Status: Offline
Reply With Quote
May 11, 2001, 10:13 AM
 
milio -

Saying it is more computationally expensive does not directly corelate to quartz being slow. I want examples of where Quartz is falling down on the job. How is quartz adversly affecting the user experience.

I just don't think quartz is the problem. It may be more complex, have more features and require more proc. power but it doesn't seem to slow the user down at all.
'Satisfy the urge and discover the need' Q-Tip
     
isidorbis
Guest
Status:
Reply With Quote
May 11, 2001, 10:51 AM
 
"2) scrolling, fast"

How are you scrolling ? if you use the arrows, it's OK. But if you drag the srollbar on a complex webpage in OmniWeb, it's very slow. Same thing if you use next page (clicking out of the scrollbar) : the delay between the click and the result is almost 1/2 second.

(Please don't say "on my G3 266, scrolling is blindly fast" before trying scrolling a big list in the finder or a big page in Omni Web).
     
ntsc
Senior User
Join Date: Sep 2000
Location: Glasgow, Scotland UK
Status: Offline
Reply With Quote
May 11, 2001, 11:17 AM
 
As a complete non-programmer i would like to know from those of you who say Quartz is not the problem and Carbon events are. Why is it that Window resizing is slow in OmniWeb which is a Cocoa browser if it is Carbon events slowing us down?

To me it would seem that while Quartz is the superior technology with respect to display architectures it is not fully matured yet but it is kind of a chicken and the egg thing because GPU manufacturers will not provide acceleration to a system that is currently vapourware so Apple had to get this out the door.

------------------
"It is easy to be brave from a safe distance"
-Aesop

iMac DV 400Mhz
Mac OS X 10.0.2
320MB RAM
"You can't waste a life hating people, because all they do is live their life, laughing, doing more evil."

-ALPHA ROBERTSON,whose daughter was one of four girls killed in the bombing of a Birmingham, Ala., church in 1963.
     
 
Thread Tools
 
Forum Links
Forum Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Top
Privacy Policy
All times are GMT -4. The time now is 09:09 PM.
All contents of these forums © 1995-2017 MacNN. All rights reserved.
Branding + Design: www.gesamtbild.com
vBulletin v.3.8.8 © 2000-2017, Jelsoft Enterprises Ltd.,